An economic analogy to thermodynamics

Wayne M. Saslow

Citation: Am. J. Phys. 67, 1239 (1999); doi: 10.1119/1.19110

View online: http://dx.doi.org/10.1119/1.19110
View Table of Contents: http://ajp.aapt.org/resource/1/AJPIAS/v67/i12
Published by the American Association of Physics Teachers

Related Articles
MicroReviews by the Book Review Editor: Beyond the Black Box: The Forensics of Airplane Crashes: George
Bibel
Phys. Teach. 50, 511 (2012)
MicroReviews by the Book Review Editor: Particle or Wave: The Evolution of the Concept of Matter in Modern
Physics: Charis Anastopoulos
Phys. Teach. 50, 511 (2012)
MicroReviews by the Book Review Editor: The Rise of Nuclear Fear: Spencer R. Weart
Phys. Teach. 50, 511 (2012)
Everyday Engineering: Putting the E in STEM Teaching and Learning: Richard H. Moyer and Susan A. Everett
Phys. Teach. 50, 511 (2012)
MicroReviews by the Book Review Editor: The Little Brown Book
Phys. Teach. 50, 511 (2012)

Additional information on Am. J. Phys.

Journal Homepage: http://ajp.aapt.org/
Journal Information: http://ajp.aapt.org/about/about_the_journal
Top downloads: http://ajp.aapt.org/most_downloaded
Information for Authors: http://ajp.dickinson.edu/Contributors/contGenInfo.html

Downloaded 19 Nov 2012 to 152.3.102.242. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission
An economic analogy to thermodynamics
Wayne M. Saslow
Department of Physics, Texas A&M University, College Station, Texas 77840
共Received 28 December 1998; accepted 12 May 1999兲
We develop analogies between economic systems and thermodynamics, and show how economic
quantities can characterize the state of an economic system in equilibrium. We argue that just as a
physical system in thermodynamic equilibrium requires a nonmechanical variable 共the temperature
T) to specify its state, so does an economic system. In addition, both systems must have a
corresponding conjugate quantity, the entropy S. We also develop economic analogies to the free
energy, Maxwell relations, and the Gibbs–Duhem relationship. Assuming that economic utility can
be measured, we develop an operational definition of an economic temperature scale. We also
develop an analogy to statistical mechanics, which leads to Gaussian fluctuations. © 1999 American
Association of Physics Teachers.

I. INTRODUCTION maximization principle that leads to Gaussian fluctuations.

We propose a theory for the phenomenological description
of economic systems, based on the analogous description of
thermodynamic systems. Like thermodynamics, the present
theory has predictive power only to the extent that it can
relate different sets of measurements. Within this context, it
makes many statements that extend beyond conventional
economic theory. Fundamental to our analysis is the assump-
tion that the economic system be in equilibrium; however, it
is not our purpose to determine when a system is in equilib-
rium.
Several years ago, Kagel and Battalio1 performed a set of
experiments on the economic behavior of rats. Price was
determined by the number of bar-presses per payoff from a
fluid dispenser. The rats were given two desirable commodi-
ties whose prices were then varied. These authors found rat
behavior to be both deterministic 共because of the well-
defined average behavior兲, and probabilistic 共because of fluc-
tuations about the average behavior兲. The observed fluctua-
tions suggest that there is an economic equivalent to the
temperature T and its conjugate variable S.
Relatively little of the economics literature discusses fluc-
tuations of a single economic system, a notable exception
being the work by Theil.2 However, there is a long tradition
in economics of considering analogies between economic
and physical systems.3 In particular, there have been studies
in economics that develop various implications of the con-
cept of entropy.4–7 However, Refs. 4–7 do not use entropy in
the same way as we do here—as a quantity that defines the
internal state of an economic system.
In Sec. II we give the basic elements of our analogy be-
tween economics and thermodynamics. To provide back-
ground for the details of the analogy, we review the formal-
ism and history of thermodynamics in Sec. III. In Sec. IV we
discuss the economic analogies to thermodynamics, and ob-
serve that the present formalism can be used to give a more
precise meaning to Marshall’s concept of surplus. Specifi-
cally, the formalism implies that an increase in the Marshal-
lian surplus is due to either increased leisure 共Veblenian sur-
plus兲 or increased efficiency 共Smithian surplus兲. In Sec. V
we discuss thermometry 共and ‘‘entrometry’’兲 in thermody-
namics, and in Sec. VI we discuss thermometry and entrom-
etry in economics. Section VII discusses an analogy to sta-
tistical mechanics, which supports our thermodynamic
analogy by an alternate argument based on a wealth-
Section VIII concludes with a brief summary and discussion.
II. ECONOMIC ANALOGIES
Examples of economic systems of interest are an indi-
vidual consumer or a small country, each of which is embed-
ded within a larger economic system. Consider an individual
consumer. A fundamental assumption in economics is that
the consumer employs a utility function U to choose to pur-
chase one good over another. For many purposes, it is suffi-
cient for the utility to be an ordinal quantity 共that is, it speci-
fies only relative ordering兲. However, to make the full
analogy to thermodynamics, we must take the utility U to be
a real number. We assume U to be given in a convenient set
of units, such as 1998 dollars, and we also assume that U is
measurable.8 The formalism we develop is falsifiable, and
can be overdetermined by a proper set of measurements, thus
providing constraints on its consistency. To explain the anal-
ogy, we begin by discussing certain fundamental relations in
economics.
First consider the measurable economic quantity known as
wealth,
W⫽␭M ⫹pN 共economics兲, 共1兲
where ␭ and M represent the value and amount of money,
and p and N represent vectors of prices and numbers of
goods. 共In principle, ␭ and M can be considered vectors if
the consumer has holdings in more than one type of cur-
rency. Or, we can lump both goods and money into the p and
N vectors.兲 Because W is conserved in transactions, as is the
total energy E of a thermodynamic system, it is tempting to
consider them to be analogous. From the viewpoint of maxi-
mization, a more natural analogy is between ⫺E and W, and
this analogy is pursued in Sec. VII, which discusses an anal-
ogy to statistical mechanics. However, from the viewpoint of
making an analogy to the dependence on the extensive ther-
modynamic variables, we will see in Sec. IV that the most
appropriate analogy is between E and U.
Economics assumes that the value of an individual con-
sumer’s money and goods is summarized by the value of U,
which typically exceeds W. The excess is known as the sur-
plus 共a term due to Alfred Marshall兲, for which we introduce
the notation ⌿ 共for ‘‘psurplus’’兲.9–11 Thus

1239 Am. J. Phys. 67 共12兲, December 1999 © 1999 American Association of Physics Teachers 1239

Downloaded 19 Nov 2012 to 152.3.102.242. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission
 Table I. Summary of the suggested analogies between thermodynamic and economic systems. Not listed are the
direct analogies between S and T.
Thermodynamics ⫺F
Economics W(wealth)
⌿⫽U⫺W 共economics兲.
In a primitive or very poor economy, there is no surplus, so
⌿⫽0. In this case, every individual performs the same eco-
nomic function at the same efficiency, and there is no benefit
from specialization and trade. The surplus ⌿ cannot be nega-
tive; for typical economic systems ⌿⬎0. Although Eq. 共2兲
appears only to define another unknown quantity, ⌿, in
terms of U, this economic relationship is useful because it
has a thermodynamic analogue.
The Helmholtz free energy of a system with N identical
particles is defined as
F⫽⫺ PV⫹ ␮ N 共thermodynamics兲,
where P is the pressure, V is the volume, and ␮ is the chemi-
cal potential of the particles. We may think of ⫺ PV as
analogous to ␭M . 共There may be a deeper relationship,
wherein the ‘‘economic engine’’ uses money ␭dM rather
than work ⫺ PdV, but we do not pursue this relationship
here.兲 The quantity in thermodynamics analogous to the
price p is the chemical potential ␮. The energy E is related
to F in terms of the temperature T and entropy S via
TS⫽E⫺F 共thermodynamics兲.
Note that, according to the third law of thermodynamics, S
⫽0 for a system at T⫽0.
A comparison of Eqs. 共2兲 and 共4兲 suggests another anal-
ogy, that of ⌿ and TS. By taking a system with zero surplus
共and thus zero economic temperature兲 to have zero economic
entropy, we assume the economic analogue of the third law
of thermodynamics.
Because the surplus is zero for an undeveloped economy,
we make the tentative common language identification of
economic temperature T with the level of economic develop-
ment. This usage seems consistent with the idea that T is an
intensive quantity. We propose no common language defini-
tion of economic entropy, although we expect it to be related
to economic variety, which in turn may be a measure of the
economic value of leisure. We summarize the thermody-
namic and economic analogies in Table I.
III. THERMODYNAMICS, ENTROPY, AND
EQUILIBRIUM
The concepts of temperature and entropy are well devel-
oped in thermodynamics. However, when thermodynamics
was in its nascent state, these concepts were obscure. To
provide background for our development of the economic
analogy to thermodynamics, we give a brief review of cer-
tain key ideas in thermodynamics.
Thermodynamics deals with the transformation of heat
into mechanical work, and dates from the work of Carnot
共1824兲, who established the principle limiting the amount of
work obtainable from heat under given conditions. The rea-
1240 Am. J. Phys., Vol. 67, No. 12, December 1999
Downloaded 19 Nov 2012 to 152.3.102.242. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission
⫺E TS ␮ N
U(utility) ⌿(surplus) p(price) N共# of goods兲
共2兲 soning by which Carnot established his principle ‘‘... is justly
regarded as one of the most remarkable triumphs of the de-
ductive method.’’ 12
Carnot was concerned with the efficiency of heat engines.
He determined that for a heat engine operating between two
nearby temperatures, the ideal efficiency 共useful work di-
vided by the heat provided by the hotter reservoir兲 is propor-
tional to the temperature difference times an unknown but
universal function of the average temperature. Carnot’s
thinking is the basis of the second law of thermodynamics. It
is remarkable that it was established well before the experi-
mental studies 共1845兲 by James Joule that established the
first law of thermodynamics that heat is a form of energy.
共3兲
According to Joule, a system with a fixed number of par-
ticles can change its energy in either of two ways: heat dQ
can enter the system or work dW can be done on the system.
Energy conservation is written as
dE⫽dQ⫹dW. 共5兲
From Carnot’s studies we can show that, in equilibrium,
dQ⫽TdS, where T is the temperature and the entropy S is a
function of the state of the system. Hence dS⫽dQ/T is a
true differential. It took 30 years to conceive and then to
共4兲 establish this simple but nontrivial statement. Because the
work done on the system is dW⫽⫺ PdV, energy conserva-
tion takes the form
dE⫽TdS⫺ PdV. 共6兲
Although the energy of a thermodynamic system is
uniquely defined, its heat content 共the integral of dQ) and its
work content 共the integral of ⫺ PdV) are not. A system can
go from one energy state to another via an infinite number of
processes where the contributions of heat and work differ.
Thus neither dQ nor dW is an exact differential. However,
dW/ P⫽⫺dV and dQ/T⫽dS are exact differentials.
Finally, if particles can enter or leave the system by an
amount dN, there is an energy change ␮ dN. The fundamen-
tal relation of thermodynamics combines these changes of
energy to yield
dE⫽TdS⫺ PdV⫹ ␮ dN. 共7兲
Thermodynamics involves two important classes of vari-
ables. We define them by an example. Consider two cham-
bers of identical gases at the same temperature, pressure, and
chemical potential. On connecting the chambers, the energy
E, volume V, and number of particles N of the combined
system will be the sum of the energies, volumes, and num-
bers of particles for the individual systems. Such variables
are called extensive. The entropy S also is an extensive vari-
able. On the other hand, the temperature T, pressure P, and
chemical potential ␮ of the combined system will be un-
changed. Such variables are called intensive.13
Another way to write Eq. 共7兲 is to consider the energy as
an extensive variable and a function of the three extensive
variables S, V, and N:
Wayne M. Saslow 1240
 E⫽E 共 S,V,N 兲 . 共8兲 Equating Eqs. 共13兲 and 共14兲 yields the fundamental relation-
ship
One goal of thermodynamics is to provide a theoretical
framework so that experimental measurements can deter- E 共 S,V,N 兲 ⫽TS⫺ PV⫹ ␮ N. 共15兲
mine this functional dependence for a given system. A Subtracting Eq. 共9兲 from the differential of Eq. 共15兲 yields
knowledge of E(S,V,N) completely characterizes the ther- the Gibbs–Duhem relation
modynamic state of the system. For this reason E(S,V,N) is
called the state function for the system. 0⫽SdT⫺Vd P⫹Nd ␮ . 共16兲
Because changes in the energy E are characterized by For some purposes, the set (T,V,N) rather than (S,V,N)
changes in S, V, and N, we have from Eq. 共8兲: is a more natural set of variables. In this case, the appropriate
thermodynamic potential is the Helmholtz free energy F.
⳵E ⳵E ⳵E
dE⫽ dS⫹ dV⫹ dN. 共9兲 Combining Eqs. 共3兲 and 共15兲 yields
⳵S ⳵V ⳵N
F⫽E⫺TS. 共17兲
A comparison of Eqs. 共7兲 and 共9兲 leads to the identifications
Using Eq. 共10兲 and Eq. 共17兲, dF satisfies
T⬅ 冉 冊
⳵E
⳵S V,N
, P⬅⫺
⳵E
⳵V 冉 冊 S,N
, ␮⬅
⳵E
⳵N冉 冊 S,V
. 共10兲 dF⫽dE⫺TdS⫺SdT⫽⫺SdT⫺ PdV⫹ ␮ dN, 共18兲
which is consistent with the natural set of variables (T,V,N).
The Maxwell relations are a consequence of the fact that Like E(S,V,N), F(T,V,N) is a state function. Note that
the order of the cross derivatives of E(S,V,N) does not mat- dF⫽0 for a system at fixed V and N in contact with a
ter. Hence, from Eqs. 共8兲 to 共10兲, we have thermal reservoir that fixes T. Further, at fixed T and N, the
⳵ 2E
⫽
⳵ 2E
⳵S⳵V ⳵V⳵S
or ⫺ 冉 冊 冉 冊
⳵P
⳵S V,N
⫽
⳵T
⳵V S,N
,
work dW⫽⫺ PdV done on the system equals dF. This re-
lation is the origin of the terminology free energy.
Equation 共17兲 is the Legendre transformation of the en-
⳵ 2E
⫽
⳵ 2E
⳵S⳵N ⳵N⳵S
or 冉 冊 冉 冊
⳵␮
⳵S V,N
⫽
⳵T
⳵N S,V
, 共11兲
ergy and enables us to go from a function with natural vari-
ables (S,V,N) to one with natural variables (T,V,N). The
variables S and T are said to be dual to one another, as are
⳵ 2E
⫽
⳵ 2E
⳵N⳵V ⳵V⳵N
or ⫺ 冉 冊 冉 冊
⳵P
⳵N S,V
⫽
⳵␮
⳵V S,N
.
the variables V and P and N and ␮. If a system has two
possible states with the same T, V, and N, the state with the
lower free energy F(T,V,N) is thermodynamically stable.
These relations guarantee that integrals over dE in S, V, N Using its natural variables, Eq. 共3兲 for F may be written as
space are path independent. F 共 T,V,N 兲 ⫽⫺ PV⫹ ␮ N. 共19兲
Because the energy is an extensive quantity, it satisfies13
By taking two cross-derivatives with respect to its variables,
E 共 ␣ S, ␣ V, ␣ N 兲 ⫽ ␣ E 共 S,V,N 兲 , 共12兲 we can derive three new Maxwell relations from the free
for arbitrary ␣ ⬎0. That is, scaling the extensive quantities S, energy F(T,V,N).
V, and N also scales the extensive quantity E, which de-
pends upon S, V, and N. Note that T, P, and ␮ are un- IV. RELATING THERMODYNAMICS AND
changed because this scaling corresponds to a system that is ECONOMICS
␣ times larger. Differentiating the left-hand side of Eq. 共12兲 Our goal in making an analogy between economics and
with respect to ␣, and then using both Eqs. 共12兲 and 共10兲, thermodynamics is to provide a theoretical framework so
gives that economics measurements can determine the functional
dE 共 ␣ S, ␣ V, ␣ N 兲 ⳵ E 共 ␣ S, ␣ V, ␣ N 兲 ⳵ 共 ␣ S 兲 dependence of the utility U on the economic parameters that
⫽ specify the state of an economic system. A knowledge of the
d␣ ⳵共 ␣S 兲 ⳵␣
state function as a function of the appropriate economic pa-
⳵ E 共 ␣ S, ␣ V, ␣ N 兲 ⳵ 共 ␣ V 兲 rameters completely characterizes the economic system.
⫹ From the economic relations introduced in Sec. I, we have
⳵共 ␣V 兲 ⳵␣
⌿⫽TS, 共20兲
⳵ E 共 ␣ S, ␣ V, ␣ N 兲 ⳵ 共 ␣ N 兲
⫹ and
⳵共 ␣N 兲 ⳵␣
U⫽TS⫹W⫽TS⫹␭M ⫹ pN. 共21兲
⳵ E 共 S,V,N 兲 ⳵ E 共 S,V,N 兲
⫽ S⫹ V A comparison of Eq. 共21兲 for U to Eq. 共15兲 for E suggests
⳵S ⳵V
that, from the point of view of its natural set of variables, we
⳵ E 共 S,V,N 兲 have
⫹ N⫽TS⫺ PV⫹ ␮ N.
⳵N U⫽U 共 S,M ,N 兲 . 共22兲
共13兲 Relation 共22兲 is our fundamental assumption.
Differentiating the right-hand side of Eq. 共12兲 gives The economic equivalent of Eq. 共7兲 is
dE 共 ␣ S, ␣ V, ␣ N 兲 dU⫽TdS⫹␭dM ⫹ pdN, 共23兲
⫽E 共 S,V,N 兲 . 共14兲
d␣ where

1241 Am. J. Phys., Vol. 67, No. 12, December 1999 Wayne M. Saslow 1241

Downloaded 19 Nov 2012 to 152.3.102.242. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission
 T⬅ 冉 冊
⳵U
⳵S M ,N
, ␭⬅ 冉 冊 ⳵U
⳵M S,N
, p⬅ 冉 冊
⳵U
⳵N S,M
. 共24兲
The ratio m/ p can be generalized to include the value of
currency, thus permitting the study of saving. Specifically,
define

冉 冊
Let us now apply this theoretical structure.
⳵U
In the The Wealth of Nations, Adam Smith distinguishes m ␭⬅ . 共31兲
between two measures of utility.14 One measure is the ⳵M x,N
‘‘value in exchange.’’ In economics it is conventional to
Then the ratio of value in use to value in exchange for
identify the value in exchange with the price p. From 共24兲,
money, m ␭ /␭, takes on the same value as m/p for goods.
we take this measure to be the marginal utility per good
dU/dN at fixed S and M . 共This last statement is deceptive, If we use Eq. 共21兲 to relate W and U, the analogy to the
because we have not yet given S an operational definition.兲 development associated with F leads to
Another measure is the ‘‘value in use,’’ which is less readily dW⫽⫺SdT⫹␭dM ⫹ pdN, 共32兲
identified.
We will identify ‘‘value in use’’ with the marginal utility where
per good dU/dN for another set of fixed variables. For sim- ⳵W ⳵W ⳵W
plicity, we will take M to be fixed, but we cannot be explicit S⬅⫺ , ␭⬅ , p⬅ . 共33兲
⳵T ⳵M ⳵N
about the second variable that is to be held fixed, and will
simply denote it as x. 共Perhaps the second variable is T; our It is implicit that two of the three quantities (T, M , N) are
uncertainty about which variable is to be held fixed is an held constant in the partial derivatives. From Eq. 共32兲 we
echo of conventional economics, where the imprecise phrase may write the functional dependence
‘‘all other quantities held constant’’ is commonly invoked.兲
From Eq. 共23兲 we then have W⫽W 共 T,M ,N 兲 . 共34兲

冉 冊
⳵U
⳵N x,M
⫽T 冉 冊 ⳵S
⳵N x,M
⫹p. 共25兲
Another standard economic relationship states that when
an individual consumer interacts with the market, the price
共marginal cost per good兲 is determined by the market. We
One of the great triumphs of nineteenth century ‘‘margin- can obtain this result by assuming that, in equilibrium, the
alist’’ economic theory is the following statement: A con- total wealth of the consumer and of the market is maximized
sumer will purchase goods subject to the condition that the at fixed temperature and money. Considering the market to
ratio of the value of any good in use to its price p takes on a be a reservoir r, we have from Eq. 共32兲
common value.8 This statement follows on requiring that, for dW r ⫽⫺S r dT r ⫹␭ r dM r ⫹ p r dN r . 共35兲
fixed market values of goods, U be maximized for each
good. Fixed market value means that the goods 1 and 2 are Subject to the conditions dT⫽dT r ⫽0, conservation of
exchanged in the marketplace subject to the condition money (dM ⫹dM r ⫽0), and conservation of goods (dN
⫹dN r ⫽0), we find by adding Eqs. 共32兲 and 共35兲 that
0⫽p 1 dN 1 ⫹p 2 dN 2 . 共26兲
dW⫹dW r ⫽ 共 ␭⫺␭ r 兲 dM ⫹ 共 p⫺ p r 兲 dN. 共36兲
The maximization of U requires that
The right-hand side of Eq. 共36兲 is zero for arbitrary varia-
⳵U ⳵U tions dM and dN only if the value of money to the consumer
0⫽ dN 1 ⫹ dN 2 . 共27兲
⳵N1 ⳵N2 is the same as the value of money to the market: ␭⫽␭ r .
Combining Eqs. 共26兲 and 共27兲 then yields Similarly, for the value of a good, we have p⫽p r .
Note that Eq. 共1兲 gives the differential
1 ⳵U
⫽constant 共28兲 dW⫽␭dM ⫹pdN⫹M d␭⫹Nd p. 共37兲
p ⳵N
The consistency of Eqs. 共37兲 and 共32兲 requires that
for each good. Thus, as desired, the ratio of value in use to
price is a constant. 0⫽SdT⫹M d␭⫹Nd p. 共38兲
Using Eq. 共25兲, Eq. 共28兲 can be expressed as
Equation 共38兲 is the analogue of Eq. 共16兲, the Gibbs–Duhem
冉 冊
1 ⳵U
p ⳵N x,M
⫽
T ⳵S
p ⳵N 冉 冊 x,M
⫹1⫽constant. 共29兲
relation. Among other things, it implies that a decrease in the
price of money or goods 共as when the state of economic
development increases兲 is accompanied by an increase in the
From Eq. 共29兲, the constancy of this ratio for all goods does economic temperature. This qualitative behavior is expected
not depend on whether the price is included in the computa- from conventional economic reasoning. Specifically, if all
tion of utility. The present formalism helps us focus on the prices and currency values are increased by a common fac-
issue of ‘‘what is held constant.’’ Let m represent the value tor, then the system does not really change: By Eq. 共38兲 the
in use 共marginal utility per good, at fixed x and M ): temperature is simply increased by the common factor. How-
ever, Eq. 共38兲 holds for more general variations. Note that
m⬅ 冉 冊
⳵U
⳵N x,M
. 共30兲
scaling by a factor of 2 means calling a one-dollar bill a
two-dollar bill, etc. To really change the value of the dollar
would require printing more dollars, which is a real cost that
Note that m is specified in monetary units. From Eqs. 共28兲 cannot be scaled away.
and 共29兲, the ratio m/p has the same dimensionless value for Equation 共38兲 has an important application. We can write
all goods. the change in the Marshallian surplus, ⌿⫽TS, as

1242 Am. J. Phys., Vol. 67, No. 12, December 1999 Wayne M. Saslow 1242

Downloaded 19 Nov 2012 to 152.3.102.242. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission
 d⌿⫽TdS⫹SdT⫽TdS⫺M d␭⫺Ndp.
We interpret the term TdS as the change in the economic
value of leisure. We shall call TdS the Veblenian surplus
共for the economist Thorstein Veblen, known for his studies
of leisure兲. As we will show shortly 共and as is well-known to
economists兲, the term ⫺Ndp is the change in the consumer’s
surplus of goods. We shall call ⫺Ndp the Smithian surplus
共for the economist Adam Smith, who argued that efficiencies
produced by specialization lead to such surpluses兲. Likewise,
the term ⫺M d␭ may be interpreted as a currency surplus,
due to efficiencies produced by specialization.15 We shall
also call this currency surplus a Smithian surplus.
To see that ⫺Nd p is a surplus, note that the cost of incre-
mentally purchasing goods 共where the first goods are scarce
and, hence, costly兲 is 兰 N0 pdN, where p(N), the price of the
Nth good, decreases as N increases 共that is, dp/dN⬍0.)
However, when purchased all at once, the actual cost to the
consumer is Np, the number of goods times the latest cost
per good. The difference is
冕 冕
N p(N)
pdN⫺Np⫽⫺ Ndp.
0 p(⫽0)
The difference is positive, because for the limits of integra-
tion in Eq. 共40兲, dp is negative. Hence ⫺Ndp is the change
in the consumer’s surplus of goods.
Equation 共39兲 shows that there are two types of surplus:
the Veblenian surplus of leisure and the Smithian surplus
from efficiency due to specialization. These ideas are present
in economics, but we are unaware of any previous formal
statement that relates the Marshallian, Veblenian, and Smith-
ian surpluses. Note that the statement of the constancy of
m/ p for all goods purchased by a given consumer is the
same as the statement that the ratio of Veblenian surplus per
good to the price per good is a constant for all goods pur-
chased by a given consumer.
The following ‘‘Maxwell relations’’ are an immediate
consequence of the fact that the order of the cross derivatives
of U⫽U(S,M ,N) does not matter. Thus
冉 冊 冉 冊
⳵ 2U ⳵ 2U ⳵p ⳵T
⫽ or ⫽ ,
⳵S⳵N ⳵N⳵S ⳵S M ,N
⳵N
冉 冊 冉 冊
⳵ 2U ⳵ 2U ⳵␭ ⳵T
⫽ or ⫽ ,
⳵S⳵M ⳵M ⳵S ⳵S M ,N
⳵M
⳵ 2U
⫽
⳵ 2U
⳵M ⳵N ⳵N⳵M
or 冉 冊 冉 冊
⳵p
⳵N S,V
⫽
⳵␭
⳵M
In economics, these relations are known as Slutsky
conditions.16,17 They guarantee that integrals over dU in S,
N, M space are path independent. Similar Slutsky conditions
can be derived from W, for which the natural variables are
T, M , and N. Equations 共41兲 and 共42兲 are new; Eq. 共43兲 is
already known in a form where the dependence on entropy is
not made explicit. The addition of the variables T and S
helps makes more precise the meaning of the phrase ‘‘all
other quantities held constant.’’
V. THERMOMETRY IN THERMODYNAMICS
We have so far assumed that economic ‘‘temperature’’ is a
well-defined quantity. However, in thermodynamics it took
1243 Am. J. Phys., Vol. 67, No. 12, December 1999
Downloaded 19 Nov 2012 to 152.3.102.242. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission
共39兲 hundreds of years before the qualitative thermometers based
upon the height of a column of fluid were properly calibrated
against an absolute temperature scale T. The ideal gas law
makes the determination of absolute temperature a relatively
simple one. When it is applicable to real gases at low density
N/V and high absolute temperature T, the ideal gas law,
PV⫽Nk B T, makes thermometry relatively easy (k B is
Boltzmann’s constant兲. Hence, a measurement of P and N/V
gives T.
It will be difficult to find an economic analogue to the
ideal gas thermometer. One difference is that the speed of the
particles of an ideal gas has no upper bound and a bounded
lower limit on its energy, whereas an economic agent has
bounded values for its wealth and utility. However, magnetic
salts behave, for a limited range of temperatures, as if they
have only a finite number of energy levels.
Common thermometry involves reading the height of a
column of fluid or the position of a pointer attached to a coil
of wire. Both of these quantities depend upon the thermal
expansion coefficient. Other forms of thermometry depend
upon other temperature-dependent variables. At room tem-
共40兲 perature, these can be calibrated against another thermometer
based on the ideal gas law. At lower temperature it is con-
venient to use the carbon resistor thermometer; its electrical
resistance is a measure of the temperature. In the millikelvin
range, calibrations are done with thermometers that use cer-
tain magnetic salts. At very low and very high temperatures,
it is difficult to perform any thermometry at all. We therefore
consider the general problem of how we can calibrate a mea-
surement of a measured quantity 共for example, electrical re-
sistance兲, which we will call ␶, against an absolute thermo-
dynamic temperature T. 18
Recall that T⫽0 for all thermodynamic temperature
scales. However, there is no absolute scale for temperature.
By setting 共approximately兲 273 deg to be the freezing point
of water, or using the temperature of the triple point of a pure
material, we determine the Kelvin scale of temperature T K .
On another planet, the thermodynamic temperature scale
would be different, but only by a scale factor. If the inhab-
itants of the other planet called the freezing point of water
546 deg, we would know that their temperatures are all twice
共41兲
S,M
as high as on the kelvin scale.
Determining the temperature scale. In general, the ther-
mometer property ␶ depends on T, P, and N. For this ther-
共42兲
S,N
mometer, we must determine the thermodynamic tempera-
ture T as a function of ␶, V, and N. Our discussion is an
S,N
. 共43兲
extension of Landau and Lifshitz,18 who consider a ther-
mometer that measures a quantity ␶ depending only upon the
temperature T: ␶ ⫽ ␶ (T).
Consider a measurable quantity, the heat gain dQ⫽TdS,
and how it varies with a change in pressure at fixed tempera-
ture T and fixed N. We will find it helpful to employ the
following Maxwell relation 关based on the Gibbs free energy
G(T, P,N)⫽E⫺TS⫹ PV]:
冉 冊 冉 冊
⳵S ⳵V
⫺ ⫽ . 共44兲
⳵P T,N
⳵T P,N
From Eq. 共44兲 we can write
Wayne M. Saslow 1243
 冉 冊
⳵Q
⳵P
⫽T 冉 冊
⳵S
⳵P
⫽⫺T 冉 冊
⳵V
⳵T
which follows on substituting Eq. 共23兲 in the differential of
Eq. 共50兲. Then

冉 冊 冉 冊
T,N T,N P,N

⫽⫺T
⳵V
⳵␶冉 冊 冉 冊 P,N
⳵␶
⳵T P,N
, 共45兲
⳵ 2V
⳵␭⳵T N
⫽
⳵ 2V
⳵T⳵␭ N
, 共52兲

which can be re-arranged to read which leads to

1 ⳵T
T ⳵␶ 冉 冊 P,N
⫽⫺
共 ⳵ V/ ⳵ ␶ 兲 P,N
共 ⳵ Q/ ⳵ P 兲 T,N
⬅ f 共 ␶ , P,N 兲 . 共46兲 冉 冊 冉 冊
⳵S
⳵␭ ␭,N
⫽
⳵M
⳵T ␭,N
. 共53兲

The left-hand side of Eq. 共46兲 is at fixed P and N, thus Using Eq. 共53兲 and dQ⬅TdS, we have
explaining the notation f ( ␶ , P,N). Equation 共46兲 tells us that
the fractional change in temperature can be expressed in
terms of the measurable quantities f ( ␶ , P,N) and d ␶ . We can
integrate Eq. 共46兲 to obtain
冉 冊
⳵Q
⳵␭ T,N
⫽T 冉 冊
⳵S
⳵␭ T,N
⫽⫺T 冉 冊
⳵M
⳵T ␭,N

ln
T
T0
⫽ 冕␶0
␶
f 共 ␶ , P,N 兲 d ␶ , f 共 ␶ , P,N 兲 ⫽⫺
共 ⳵ V/ ⳵ ␶ 兲 P,N
共 ⳵ Q/ ⳵ P 兲 T,N
, ⫽⫺T 冉 冊 冉 冊
⳵M
⳵␶ ␭,N
⳵␶
⳵T ␭,N
, 共54兲

共47兲 which can be re-arranged to read

where dQ⬅TdS. In this way we can obtain the temperature
scale T as T⫽T( ␶ , P,N) in terms of measurable quantities.
By using fixed V and varying N, we can obtain T( ␶ ,V, ␮ )
1 ⳵T
T ⳵␶ 冉 冊 ␭,N
⫽
共 ⳵ M / ⳵ ␶ 兲 ␭,N
共 ⳵ Q/ ⳵ ␭ 兲 T,N
⬅ f 共 ␶ ,␭,N 兲 . 共55兲

via the relations The left-hand side of Eq. 共55兲 is at fixed ␭ and N, so that

冕
T ␶
ln ⫽ g 共 ␶ ,V, ␮ 兲 d ␶ , g 共 ␶ ,V, ␮ 兲 ⬅
T0 ␶0
Thus there is more than one way to obtain a temperature
scale. Equations 共47兲 and 共48兲 must be consistent with one
another.
The above discussion suggests at least eight ways to per-
form thermometry. At fixed T, we measure the dependence
of dQ on either dV or d P, and at either fixed N or fixed ␮,
we measure the dependence of dQ on either dN or d ␮ at
either fixed V or fixed P. Two pairs of these eight measure-
ments lead to ␶ as a function of the same fixed variables.
VI. THERMOMETRY IN ECONOMICS
Having described some of the complexity of thermometry
in physics, we now indicate how to perform thermometry in
economics. To do so, it is essential that we be able to mea-
sure dQ⬅TdS, the Veblenian surplus, and that we have a
quantity that can serve as a thermometer. By Eq. 共23兲, we
have
共 ⳵ N/ ⳵ ␶ 兲 ␮ ,V
. ␶, ␭, and N, thus explaining the notation f ( ␶ ,␭,N). We can
共 ⳵ Q/ ⳵ ␮ 兲 T,V
共48兲
the right-hand side must be written in terms of the variables
then integrate from T 0 and the corresponding ␶ 0 to obtain
ln
T
T0
⫽ 冕␶0
␶
f 共 ␶ ,␭,N 兲 d ␶ , f 共 ␶ ,␭,N 兲 ⫽
共 ⳵ M / ⳵ ␶ 兲 ␭,N
共 ⳵ Q/ ⳵ ␭ 兲 T,N
,
共56兲
where dQ⬅TdS as usual. From Eq. 共56兲 we can obtain, in
principle, the temperature scale T( ␶ ,␭,N) in terms of the
measurable quantities ␶, ␭, N. We need to be able to measure
the quantities appearing in the expression for f ( ␶ ,␭,N). Es-
pecially difficult to measure is the change in the Veblenian
surplus, dQ⬅TdS. By Eq. 共24兲, this quantity is also the
change in utility at fixed goods and money, and in that form
its measurability is the subject of debate among
economists.19 The present work is predicated upon the as-
sumption of the measurability of utility. Other forms of eco-
nomic thermometry that are perhaps more practical to imple-
ment are possible by analogy to thermometry in
thermodynamics.
VII. AN ANALOGY TO STATISTICAL MECHANICS

TdS⫽dU⫺␭dM ⫺pdN. 共49兲 In statistical mechanics we weight each microstate s with

energy E s of a system in equilibrium with a heat reservoir by
Hence, if we can measure dU, ␭, dM , p, and dN, then we
the factor exp(⫺Es /T). This weight has the following desir-
can measure dQ⬅TdS. able properties: the probabilities for independent systems are
We will assume that we have some economic indicator ␶ multiplicative; and the energy of independent systems is both
that depends upon T, ␭, and N. To employ the relation additive and conserved in interactions between the system
␶ (T,␭,N), we will require a new thermal Slutsky condition, and the reservoir. In this way we obtain the partition function
obtained using the economic analogue of the Gibbs free en-
ergy. We write Z⫽ 兺
states
e ⫺E s /T . 共57兲
V共 T,p,M 兲 ⫽U⫺TS⫺␭M , 共50兲
From Eq. 共57兲, the free energy F is defined via
where V⫽pN is obtained by using Eq. 共21兲 and is the mon-
etary value of goods. Its differential satisfies Z⫽e ⫺F/T , 共58兲
dV⫽⫺SdT⫺M d␭⫹pdN, 共51兲 so that

1244 Am. J. Phys., Vol. 67, No. 12, December 1999 Wayne M. Saslow 1244

Downloaded 19 Nov 2012 to 152.3.102.242. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission
 F⫽⫺T ln Z.
A knowledge of F yields the thermodynamic properties of
the physical system.
To make the analogy to economics, it is important to note
that the number of economic states is bounded from above,
just as for certain magnetic salts. Moreover, an economic
agent will seek to maximize utility or wealth 共according to
what variables are held constant兲, whereas a physical system
will seek to minimize energy or free energy. Four possible
weightings of economic states suggest themselves: 共1兲
e ⫺U/T ; 共2兲 e U/T ; 共3兲 e ⫺W/T ; and 共4兲 e W/T . If the relationship
U⫽W⫹TS is to be maintained, then e U/T and e ⫺W/T lead to
undesirable definitions whereby entropy is negative. Of the
remaining two weightings, e ⫺U/T would appear to be the
most natural, because of the thermodynamic analogy. How-
ever, e ⫺U/T is undesirable because it favors states of low
utility 共a minimization principle兲 and employs a quantity that
is not conserved in economic exchanges. On the other hand,
weighting by e W/T favors states of higher wealth 共a maximi-
zation principle兲 and employs a quantity that is conserved in
economic exchanges. This weight also is consistent with the
idea that T is a measure of economic development. Low
temperature 共less developed兲 economies strive for high W
states whereas high temperature 共highly developed兲 econo-
mies prefer variety, that is, economic states are equally
weighted.
Consider an economics microstate s with wealth W s . We
define the partition function
Z⫽ 兺 e W s /T ,
states
and the utility U by
Z⫽e U/T ,
so that
U⫽T ln Z.
Making the usual assumption that the sum is dominated by
the most probable states 共at the most probable wealth兲, Eq.
共62兲 becomes
Z⬇⌫e W/T , ⌫⫽ 兺
states
1,
where the sum is over economic states with wealth W s
⫽W. Combining Eqs. 共60兲–共63兲 yields
U⫽W⫹T ln ⌫.
Hence, to agree with 共2兲 and ⌿⫽TS, we identify
S⫽ln ⌫.
Equation 共65兲 is the economic analogue of the famous rela-
tionship due to Boltzmann. It relates economic entropy to
economic variety.
We will not explore how to count economic states, which
is a complex and difficult subject, involving considerations
as complex as the determination of the ‘‘phase space’’ asso-
1245 Am. J. Phys., Vol. 67, No. 12, December 1999
Downloaded 19 Nov 2012 to 152.3.102.242. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission
共59兲 ciated with leisure. 共Recall the difficulties associated with
state counting in physics before quantum mechanics defined
the unit of phase space.兲
From F⫽E⫺TS, low temperature favors E minimization,
whereas high temperature favors S maximization. Corre-
spondingly, from U⫽W⫹TS, low economic temperature fa-
vors W maximization, whereas high economic temperature
favors S maximization. From the viewpoint of maximization,
⫺F and U are analogous. This analogy is in contrast to our
earlier discussion, where E and U are analogous from the
viewpoint of natural variables. Note that U calculated from
Eq. 共62兲 will be written in terms of T, which is not its natural
variable. By use of W⫽U⫺TS, where S⫽ 兰 T0 (dT/T)
⫻( ⳵ U/ ⳵ T) 兩 M ,N , the function W(T,M ,N) can be obtained,
from which the economics can be calculated.
Equation 共65兲 can be made the basis of a theory of fluc-
tuations about economic equilibrium in analogy to the theory
of fluctuations about thermodynamic equilibrium. For ex-
ample, the fluctuations ␦ N in the number of particles N
satisfy18
⳵N
共 ␦ N 兲 2 ⫽T , 共66兲
⳵␮
where T has the same units 共energy兲 as ␮. Analogously, the
fluctuations ␦ N in the number of goods N have a mean
square average which satisfies
⳵N
共 ␦ N 兲 2 ⫽⫺T , 共67兲
⳵p
共60兲
where T has the same units 共for example, 1998 dollars兲 as p.
The derivative in Eq. 共67兲, which appeared in our earlier
discussion of Eq. 共40兲, is proportional to the price-elasticity
(p/N) ⳵ N/ ⳵ p. Theil2 assumed that, for a given economic
共61兲
agent, each good will satisfy a relationship similar to Eq.
共67兲. Although he did not give the coefficient of proportion-
ality, his economic intuition led him to suggest that the co-
efficient of proportionality should be the same for all goods.
共62兲
共He did not include the concept of temperature in this con-
text.兲 Equation 共67兲 goes beyond Thiel in predicting that the
fluctuations increase for fixed T, if ⳵ N/ ⳵ p 共akin to a suscep-
tibility or compressibility兲 increases, and that the fluctuations
increase as T increases, for fixed ⳵ N/ ⳵ p. Note that if ( ␦ N) 2
and ⳵ N/ ⳵ p can be measured for an equilibrium system with
共63兲 Gaussian fluctuations, then Eq. 共67兲 could be used to deter-
mine the economic temperature.
VIII. SUMMARY AND DISCUSSION
共64兲 Assuming that the utility U can be given a dollar value, so
that economic thermometry can be performed 共in principle兲,
we have developed an analogy between the economic quan-
tities surplus ⌿, utility U, and wealth W, and the thermody-
共65兲 namic quantities TS, energy E, and Helmholtz free energy
F. In Sec. II we tentatively made a common language iden-
tification of T with the level of economic development, but
we have avoided a common language identification of S. The
quantities T and S correspond, at least in part, to psychologi-
cal variables, which often are invoked in economic discus-
sions of utility.
Wayne M. Saslow 1245
 This analogy has lead to a number of apparently new re-
sults: an economic Gibbs–Duhem relation; the equating of
Marshallian surplus ⌿ to TS; the use of the economic
Gibbs–Duhem relation to recast the Smithian surplus
⫺Ndp⫺M d␭ as SdT; the equating of TdS to increase in
utility at fixed goods N and money M ; the interpretation of
TdS as Veblenian surplus; new and more precise Slutsky
relations; and the relationship between measurements of util-
ity and the establishment of an economic temperature scale.
We also developed an analogy to the relation between statis-
tical physics and thermodynamics. Moreover, Eq. 共65兲 can,
in principle, be used to study near-equilibrium fluctuations.
In the late nineteenth century, Irving Fisher3 developed a
mechanical analogy between economics and physics, invok-
ing force and distance to be analogous to price and number
of goods, respectively. In the present work we have made a
thermodynamic analogy, invoking the chemical potential and
number of goods, and especially the temperature and en-
tropy. Such a viewpoint would have been recognized imme-
diately by Fisher’s advisor—J. Willard Gibbs himself.
Neither we nor professional economists know if the as-
sumption of economic equilibrium is correct. Furthermore,
we do not know if econometric measurements are accurate
enough to test the theory, even if a system is found that is
thought to be in equilibrium. Nevertheless, even if no eco-
nomic system is found to be in true equilibrium, the analo-
gies we have proposed here may have value for systems that
are slightly out of equilibrium.
A surprisingly large number of apparently new results
have been obtained using the macroscopic reasoning associ-
ated with thermodynamics. Of course, it is possible that a
better analogy can be developed. We assume economic sys-
tems that are permanent and are in equilibrium and that
wealth is conserved under exchange with fixed prices. These
assumptions cannot be literally true. Moreover, although the
formalism allows for an internal variable 共the temperature兲,
this value has been thought of as a societally determined
quantity 共the state of economic development兲, rather than a
physiological-psychological variable associated with the in-
dividual. Probably the best way to test the economic-
thermodynamic analogy is by applying it to economic mea-
surements on systems that are thought to be in, or nearly in,
economic equilibrium. Note that economic individuality will
appear in the micro-aspects of the economic-statistical-
mechanics analogy of Sec. VII. Probably the best way to test
this analogy is by developing microscopic models for the
behavior of specific economic systems. Even if the analogies
require modification, they suggest directions for future re-
search.
Suggested problems: The interested reader should consult
a basic economics text, such as Ref. 8, and a graduate text
such as Ref. 10. Reference 3 discusses various analogies that
have been made to physics.
共1兲 Develop a theory of near-equilibrium fluctuations. Note
that economic fluctuations are often non-Gaussian.20
共2兲 Develop an analogy to heat engines operating at differ-
ent temperatures. To what extent can ␭dM be consid-
ered the economic equivalent of work ⫺ PdV?
共3兲 Develop an analogy to a thermodynamic system with a
surface and surface tension. In the economic system,
land becomes a commodity with a certain value that is
analogous to surface area.
1246 Am. J. Phys., Vol. 67, No. 12, December 1999
Downloaded 19 Nov 2012 to 152.3.102.242. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission
共4兲 Develop an alternate form of economic thermometry,
where ␶ ⫽ ␶ (T,M ,N), in contrast to ␶ ⫽ ␶ (T,␭,N) of Eq.
共55兲.
共5兲 Develop a statistical mechanical model for the economic
behavior of a rat with a fixed daily number of bar presses
and a choice between two fluids with the same nutri-
tional content but different flavors. Consider the fluctua-
tions in the number of bar presses. How does price effect
the fluctuations? How does temperature effect the fluc-
tuations?
ACKNOWLEDGMENTS
I would like to thank John Kagel and Ray Battalio for
introducing me to their work on fluctuations in experimental
economics. I am indebted to Duncan Foley for suggesting
that I develop an analogy between economics and thermody-
namics. I also thank him for reading numerous drafts, for
providing focus on economics, and for general encourage-
ment. Glenn Agnolet provided valuable discussions of ther-
mometry in physical systems. Chia-Ren Hu made numerous
critical and clarifying observations. Finally, I would like to
thank the editors of this special issue for insisting on con-
ciseness and clarity, and for providing a home for an unor-
thodox application of thermodynamic and statistical-
mechanical reasoning.
1
J. H. Kagel, R. C. Battalio, L. Green, and H. Rachlin, ‘‘Consumer demand
theory applied to choice behavior of rats,’’ in Limits to Action: The Allo-
cation of Individual Behavior, edited by J. E. R. Staddon 共Academic, New
York, 1980兲. A summary of this work and others may be found in J. H.
Kagel, R. C. Battlio, and L. Green, Economic Choice Theory: An Experi-
mental Analysis of Animal Behavior 共Cambridge U.P., Cambridge, 1995兲.
2
H. Theil, Theory and Measurement of Consumer Demand 共American
Elsevier, New York, 1975兲, Vol. 1.
3
See the critical discussion of this neoclassical approach in P. Mirowski,
More Heat Than Light 共Cambridge U.P., Cambridge, 1989兲.
4
N. Georgescu-Roegen, The Entropy Law and the Economic Process 共Har-
vard U.P., Cambridge, MA, 1971兲.
5
J. Herniter, An Entropy Model of Brand Purchase Behavior 共Marketing
Science Institute, Cambridge, MA, 1972兲.
6
D. F. Batten, Spatial Analysis of Interacting Economies: The Role of En-
tropy and Information Theory in Spatial-Input Modelling 共Kluwer-Nijhoff,
Boston, 1983兲.
7
D. K. Foley, ‘‘A statistical equilibrium theory of markets,’’ J. Econ.
Theory 62, 321–345 共1994兲.
8
In discussing utility, the expression ‘‘psychological utility’’ is employed
by the classic text by P. Samuelson, Economics 共McGraw–Hill, New
York, 1970兲, 8th ed., p. 410. It is not obvious how to measure utility, but
it is believed that relative measures of utility have meaning. For some
purposes economists take a cautious position on utility, treating it as being
subject only to priority-ordering 共ordinal values兲. For other purposes
economists make the leap of faith that utility can be given a dollar value,
so that more quantitative theories can be constructed.
9
See p. 417 of Ref. 8.
10
E. Silberberg, The Structure of Economics: A Mathematical Analysis
共McGraw–Hill, New York, 1990兲, p. 396.
11
E. K. Browning and J. M. Browning, Microeconomic Theory and Appli-
cations 共HarperCollins, New York, 1992兲, 4th ed., p. 87.
12
W. Benton, Encyclopedia Brittanica 共Chicago, 1966兲, Vol. 22, p. 92.
13
H. B. Callen, Thermodynamics and an Introduction to Thermostatics
共Wiley, New York, 1985兲.
14
Adam Smith, An Inquiry into the Nature and Causes of the Wealth of
Nations, published in Dublin, printed for Messrs. Whitestone and others
共1776兲. This work can be found at http://www.bibliomania.com/
NonFiction/Smith/Wealth. See Chapter IV, ‘‘Of the Origin and Use of
Money.’’
15
Some economies employ large rocks as currency; others employ cows;
others gold; others paper. Clearly, the cost of money is not an invariant.
Wayne M. Saslow 1246
 17
Moreover, gold had a tendency to become debased. Complex printed notes A. Barten, ‘‘Evidence on the Slutsky conditions for demand equations,’’
are subject to the debasement known as counterfeiting. One can imagine Rev. Econ. Statistics 49, 77–84 共1967兲.
that there has been debasement associated with cows and large rocks. 18
L. D. Landau and E. M. Lifshitz, Statistical Physics 共Addison–Wesley,
Clearly, the formalism, as in Eq. 共39兲, must permit the value of currency ␭ Reading, MA, 1969兲, 2nd ed.
19
to vary. See p. 301 of Ref. 10.
16 20
W. E. Diewert, ‘‘Generalized Slutsky conditions for aggregate consumer B. B. Mandelbrot, Fractals and Scaling in Finance: Discontinuity, Con-
demand functions,’’ J. Econ. Theory 15, 353–362 共1977兲. centration, Risk 共Springer-Verlag, New York, 1997兲.

PHYSICS ENVY
Other scientists at the meeting said they believed that population genetics is a robust science,
but that scientists studying molecular evolution must accept the nature of their research. ‘‘We are
really about documenting patterns and trying to explain those patterns. A lot of biology is inher-
ently descriptive,’’ Dr. Staton said.
‘‘We all have physics envy,’’ said Rollin Richmond, a former student of Dobzhansky’s and
provost at the State University of New York at Stony Brook. ‘‘We can’t prove something the way
a physicist proves an electron exists. You have to have a tolerance for ambiguity.’’

David L. Wheeler, The Chronicle of Higher Education, 14 February 1997.

1247 Am. J. Phys., Vol. 67, No. 12, December 1999 Wayne M. Saslow 1247

Downloaded 19 Nov 2012 to 152.3.102.242. Redistribution subject to AAPT license or copyright; see http://ajp.aapt.org/authors/copyright_permission